Bridging the Domain Gap in Satellite Pose Estimation: a Self-Training Approach based on Geometrical Constraints

TL;DR

This paper introduces a self-training domain adaptation method for satellite pose estimation that uses geometrical constraints and segmentation to improve accuracy across different datasets, achieving state-of-the-art results.

Contribution

The authors propose a novel self-training framework leveraging geometrical constraints and segmentation to adapt satellite pose estimation models to new domains without extensive annotations.

Findings

Achieved effective domain adaptation in satellite pose estimation.

Won 1st place in the Satellite Pose Estimation Competition.

Demonstrated robustness and accuracy across different datasets.

Abstract

Recently, unsupervised domain adaptation in satellite pose estimation has gained increasing attention, aiming at alleviating the annotation cost for training deep models. To this end, we propose a self-training framework based on the domain-agnostic geometrical constraints. Specifically, we train a neural network to predict the 2D keypoints of a satellite and then use PnP to estimate the pose. The poses of target samples are regarded as latent variables to formulate the task as a minimization problem. Furthermore, we leverage fine-grained segmentation to tackle the information loss issue caused by abstracting the satellite as sparse keypoints. Finally, we iteratively solve the minimization problem in two steps: pseudo-label generation and network training. Experimental results show that our method adapts well to the target domain. Moreover, our method won the 1st place on the sunlamp task of the second international Satellite Pose Estimation Competition.